Vol. 11, No. 3MOLECULAR AND CELLULAR BIOLOGY, Mar. 1991, p. 1214-12210270-7306/91/031214-08$02.00/0Copyright 0 1991, American Society for Microbiology

A Stem Cell-Specific Silencer in the Primer-BindingSite of a Retrovirus

RICHARD PETERSEN, GERALDINE kEMPLER, AND ERIC BARKLIS*

Vollum Institute for Advanced Biomedical Research and Department of Microbiology and Immunology,Oregon Health Sciences University, Portland, Oregon 97201

Received 13 September 1990/Accepted 29 November 1990

Retrovirus expression in embryonal carcinoma (EC) cells is blocked at a postintegration stage of the viral lifecycle, in part because of the inadequate function of the viral long terminal repeat promoter in this cell type.However, selection for retrovirus expression in EC cells has identified mutations in Moloney murine leukemiavirus (M-MuLV) located in the tRNA primer-binding site (PBS) region which relieve the EC cell-specificrepression. We have found that exchanging the M-MuLV proline PBS for a glutamine one in a recombinantvirus permits expression in EC cells. By using the recombinant virus as a backbone, the EC cell-specificrepressor-binding site (RBS) element has been mapped to M-MuLV nucleotides 147 to 174. The RBS does notrequire precise positioning downstream of the M-MuLV promoter and can function in either orientation andin an intron, indicating that the regulatory effect is probably at the DNA, rather than RNA, level. We also showthat the RBS element can repress heterologous promoters from an upstream position. Our results indicate thatthe RBS acts as a silencer that its inhibitory effect is mediated by a trans-acting factor, and that the mechanismof action is probably at the level of transcription. Through in vitro binding assays we have identified a bindingfactor which specifically recognizes the wild-type RBS sequence (binding factor A). The binding characteristicsof factor A suggest that it is a stem cell repressor which acts at the M-MuLV RBS. Our DNA-binding assaysalso have identified a unique binding factor (binding factor Hp) which specifically recognizes a hemimethylatedform of the wild-type RBS. This factor may play a role in methylation mediated control of retrovirus expres-sion in EC cells.

Embryonal carcinoma (EC) cells derive from spontaneousgonadal tumors or from tumors induced by ectopic place-ment of preimplantation embryos and are the earliest cultur-able stem cells in mammalian organisms (6, 34). Spontane-ous lines develop from malignant stem cells present interatocarcinomas and display morphological, biochemical,and biological properties of pluripotent cells of the earlyembryo (6, 10). A number of differentiated cell functions arerepressed in EC cells. The inactivity of a variety of promot-ers in undifferentiated EC cells demonstrates that, at leastpartially, this regulation occurs at the level of transcription,although it is clear that inactive transcriptional units ulti-mately become repressed by DNA methylation (28, 35).One characteristic common to EC cells and preimplanta-

tion embryos is that they are nonpermissive for the expres-sion of viral genomes, including the Moloney murine leuke-mia virus (M-MuLV) genome (15, 28, 35, 36). Retroviralintegration is unimpaired in EC cells, but viral RNA accu-mulation is blocked (12, 35). The restriction of retrovirusexpression in undifferentiated EC cells appears to be primar-ily at the level of transcription (12, 20), although posttran-scriptional mechanisms have also been proposed (35). Thepromoter in the retroviral long terminal repeat (LTR) func-tions inadequately in EC cells and is at least partly respon-sible for the low level of virus-specific RNA detected (13,20). It is not clear whether the enhancer is inactive becauseof the presence of an enhancer-specific repressor, the ab-sence of specific enhancer-binding factors in EC cells, orboth.

In addition to the enhancer element, the M-MuLV provi-rus contains at least two cis-acting negative regulatory

* Corresponding author.

elements (11), one in the LTR U3 region (13) and one in thevicinity of the primer-binding site (PBS) (5, 21, 22). Theseelements inhibit virus expression specifically in undifferen-tiated cells. The PBS region, which binds a cellular tRNA toprimer first-strand synthesis during reverse transcription,appears to have a second function in mediating stem celltranscriptional repression. Loh et al. (22) have shown thatthe PBS-mediated EC cell-specific repression can be com-pletely outcompeted in DNA transfection experiments. Mu-tations in the PBS region have been shown to relieverepression in transfections as well as infections (5, 38), againindicating that the effect is not at the level of viral integra-tion. These results suggest that the tRNA PBS functionduring reverse transcription and the repressor-binding site(RBS) function, which inhibits viral expression specificallyin EC cells, are independent phenomena mediated by thesame or partially overlapping sequences.To date, the smallest sequence change to abolish the RBS

effect is the B2 mutation (5), a single-base-pair change (G toA) at M-MuLV nucleotide (nt) 160. However, owing to themechanism of retroviral replication, the B2 mutation revertsat high frequency and cannot be used to map the RBS regionin infection assays. To circumvent this problem, we havegenerated a recombinant retrovirus, PBS-glutamine (PBSQ).In PBSQ, the wild-type (wt) M-MuLV proline PBS has beenexchanged for a homologous glutamine PBS from an endog-enous provirus (7). Because PBSQ contains five single-base-pair changes from the wt M-MuLV sequence (including theB2 G-to-A mutation) and is entirely homologous to a murinetRNAGln, it is expressed in EC cells and does not revert athigh frequency. We have used M-MuLV retroviral con-structs containing PBSQ to test the effect of wt and mutantrepressor fragments on viral expression in transfections andinfections of EC cells. Our results substantiate the view that

1214

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/m

cb o

n 24

Jan

uary

202

2 by

119

.236

.85.

87.

STEM CELL-SPECIFIC SILENCER 1215

the inhibitory effect of this cis-acting element is due to thebinding of a trans-acting regulatory factor to the RBS, whichprobably limits viral expression by blocking transcription.

MATERIALS AND METHODS

Recombinant plasmids and DNA methodology. Unless oth-erwise noted, all cloning methods used were essentially asdescribed by Maniatis et al. (24). All M-MuLV designationsrefer to the M-MuLV genomic RNA as described by Shin-nick et al. (32). MP10 is a 4,670-bp provirus or a 9,200-bpplasmid and has been described by Barklis et al. (5). PBSQis identical to MP10 except that the PBS sequences havebeen exchanged for homologous sequences from an endog-enous mouse retrovirus originally isolated by Colicelli andGoff (7). The substitution of sequences from M-MuLV nt 32to 212 changes to tRNA PBS from PBS-proline in MP10 toPBS-glutamine in PBSQ. The various PBSQ constructscontain either the 47-bp (M-MuLV nt 139 to 185) or the 28-bp(M-MuLV nt 147 to 174) wt or B2 mutant (single-base-pairG-to-A mutations at M-MuLV nt 160) (5) PBS sequencesinserted into the unique BamHI site of PBSQ after beingsubcloned into identical Bluescript vectors (Stratagene).LJ-P and LJ-Q are similar to the previously described vectorDOL (18), except that portions of the 5' LTRs through theBamHI site derive from MP10 and PBSQ, respectively.LJ-PEnh- and LJ-QEnh- are variants of LJ-P and LJ-Qwith 3' LTR enhancer deletions (8) corresponding toM-MuLV nt 7938 to 8114. LJ-PAdMLPEnh- and LJ-QAdMLPEnh- are identical to LJ-PEnh- and LJ-QEnh-, exceptthat they use the adenovirus major late promoter (AdMLP,nt -248 to +34) as their internal promoter.

Cell culture and RNA analysis. EC cell lines F9 and PCC4(6), NIH 3T3 fibroblasts, and Psi2 (25) and PA317 (27)packaging cell lines were grown as described previously (5).Psi2 packaging cell populations expressing most proviruseswere generated by the transfection-infection protocol de-scribed by Jones et al. (17), and G418 selections were asdescribed previously (5). Transfections were performed bythe procedure of Graham and Van der Eb (14) as modified byParker and Stark (29), and direct-transfection protocols wereused for generation of Psi2 populations expressing enhancerdeletion constructs. Virus infection and titering, as well asRNA isolation, blotting, and hybridizations, were performedas described previously (4, 5, 38).

Protein extracts and gel shift assays. Nuclear extracts wereprepared from tissue culture cells essentially as described byDignam et al. (9) with the modifications of Baeuerle andBaltimore (2, 3), which permit the fractionation of nuclear,cytosolic, and postnuclear protein extracts. All extractswere dialyzed against Dignam buffer D (20 mM N-2-hydroxy-ethylpiperazine-N'-2-ethanesulfonic acid [HEPES; pH 7.9],100 mM KCl, 20% [vol/vol] glycerol, 0.2 mM EDTA, 0.5 mMphenylmethylsulfonyl fluoride, 0.5 mM dithiothreitol). Thepreparation of radioactive probes and the gel shift procedurewere essentially those of Thornell et al. (37), with thefollowing exceptions. The binding reactions were carried outin a total reaction volume of 20 ,ul containing 5 mM NaCl and15 to 20 mM KCl in addition to the standard Thornell bindingbuffer (25 mM HEPES [pH 7.9], 1 mM EDTA, 10% [vol/vol]glycerol, 5 mM dithiothreitol, 0.5 mM phenylmethylsulfonylfluoride), 20,000 cpm of probe (approximately 0.2 ng), and100 ng of poly(dI-dC), with a final protein concentration ofapproximately 250 ,ug/ml. Binding reactions were incubatedfor 15 to 30 min at room temperature. The electrophoresissystem used was the Tris-glycine system as described by

Thornell et al. (37). Specific double-stranded competitorswere prepared by annealing single-stranded oligonucleotidesas described by Speck and Baltimore (33). Probes were madeby kinase labeling of single-stranded oligonucleotides, an-nealing to the complementary strand, and gel purification ofdouble-stranded forms (33, 37).

Methylated probes were prepared by incubating more than106 cpm (approximately 10 ng) of wt 28-bp probe in 10 to 20U of bacterial HpaII or MspI methylase overnight at 37°C (asdescribed in the technical data from New England Biolabs).The wt 28-bp RBS of M-MuLV contains a 5'-CCGG-3'sequence which is methylated by these two methylases;HpaII methylase attaches a methyl group to the internal C,whereas MspI methylase attaches a methyl group to theexternal C. Following phenol-chloroform purification andethanol precipitation, the quality of methylation waschecked by restriction digest with methylation-sensitiverestriction enzymes and analyzed by electrophoresis on 10%native acrylamide gels. The hemimethylated forms of the wtprobes were prepared from the completely methylatedforms. Following denaturation by heating, a 10-fold molarexcess of the single-stranded oligonucleotide complemen-tary to the strand which was labeled in the preparation of theoriginal probe was added to the mixture and allowed toanneal by incubation at 50°C for 30 min, followed by a30-min incubation at room temperature. The resultingprobes, with one strand end labeled with _y32P and methyl-ated at the described nucleotide, and the other strand neitherlabeled nor methylated, were gel purified.

RESULTS

EC cell restriction maps to an RBS at M-MuLV nt 147 to174. To analyze the mechanism by which retrovirus mutantspermit EC cell expression, we have constructed a M-MuLVrecombinant, PBSQ. PBSQ is similar to its wt parent, MP10,in that its neomycin (neo) gene is expressed from theM-MuLV LTR promoter and is positioned downstream froma small intron which contains a unique BamHI cloning site(Fig. 1). However, PBSQ differs from MP10 in that wtM-MuLV nt 32 to 212 have been exchanged for homologoussequences from an endogenous mouse retrovirus (7). Thisswitch of 5 bp changes the PBSQ PBS from proline toglutamine and includes the B2 G-to-A host range mutation(5). PBSQ and MP10 are expressed very differently in ECcells relative to differentiated cells such as NIH 3T3 fibro-blasts (Table 1). With MP10, the EC cell restriction index,defined as the ratio of NIH 3T3 to EC cell titers, is greaterthan 2,000, indicating a high level of restriction. In contrast,PBSQ demonstrates a 50- to 100-fold-lower ratio, indicatingthat repression has been alleviated. It should be noted thatwe have included the results from two different PBSQ viralsupernatants to reconfirm our previous results (5, 38) thatabsolute titers (as defined by NIH 3T3 titers) may vary as aresult of factors such as the time of supernatant collectionand density of virus-producing cells, but the EC restrictionindex is relatively invariant. An additional point is thatdespite its relatively high level of expression in EC cells,PBSQ expression is reduced in this cell type relative to NIH3T3 cells. We assume that this reduction reflects the fact thatin addition to possessing a transcriptional repressor, stemcells lack positive-acting factors necessary for full activity ofthe M-MuLV LTR promoter (20).With the parental vector PBSQ, we have found it possible

to distinguish the M-MuLV leader sequence PBS functionfrom its stem cell RBS function. To map the RBS sequences

VOL. 11, 1991

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/m

cb o

n 24

Jan

uary

202

2 by

119

.236

.85.

87.

1216 PETERSEN ET AL.

1000 bp

LTR sd s NEOMP1O I-1 I II I

LTRL-P

Bamrn

PBSQ I X Ii l m

IQiw711 1IPBSO-wt47+ PBSO-wt28+

PBSO-wt47- 4-

PBSQ-m47+ *

PBSO-m47- 4

*- PBSQ-wt28-

_.o.. PBSQ-m28+

+4E~ PBSQ-m28-WT28+: 5' GGGGGCTCGTCCGGGATCGGGAGACCCC 3M28+ :5'7GG7CT1 TcCGGATCGOAGACCCC 3'

LJ-

LJ-PEnh-

LJ-QEnh-

LJ-PAdMLPEnh-

LJ-QAdMLPEnh-

BamH Hindil

LTR SV40 NEO LTR

BamH HidiI

LTR SV40 NEO LTR

BamH H tdll

LTR SV40 NEO LTR

BamHi Hidil

LTR SV40 NEO LTR

BamH Hrmdll

LTR AMUL NEO LTR

BamI Hfidil

LTR AdMLP NEO LTR

FIG. 1. Recombinant viral constructs. The previously described (5) MP10 retroviral vector expresses the neomycin (neo) gene from theM-MuLV LTR promoter. The neo gene in this vector is positioned downstream from a small intron (indicated by sd [splice donor] and sa[splice acceptor] which contains a unique BamHI cloning site. PBSQ is identical to MP10, except that RBS region sequences have beenexchanged for homologous sequences from an endogenous mouse retrovirus (7), indicated in black on the PBSQ map. Substitution ofsequences from M-MuLV nt 32 to 212 changes the tRNA PBS from PBS-proline to PBS-glutamine (PBSQ). PBSQ-wt47+, PBSQ-wt47-,PBSQ-wt28+, PBSQ-wt28-, PBSQ-m47+, PBSQ-m47-, PBSQ-m28+, and PBSQ-m28- all derive from PBSQ but contain wt or mutant m RBSfragments in either the sense (+) or antisense (-) orientation cloned into the PBSQ BamHI site. The wt 47-bp fragment contains M-MuLVDNA from nt 139 to 185; the wt 28-mer contains M-MuLV DNA from nt 147 to 174. Mutant 47- and 28-bp fragments are identical to the wtfragments except that they contain the G-to-A mutation (originally called B2 [5]) at nt 160. Sense strands of wt and mutant 28-bp fragmentsare as shown. LJ-P and LJ-Q are identical to the previously described vector DOL (18), except that portions from the 5' LTRs through theBamHI sites derive from MP10 and PBSQ, respectively. In these vectors, the neo genes are located immediately downstream from theinternal SV40 promoters (SV40 nt 208 to 5107) ( El ) LJ-PEnh- and LJ-QEnh- are variants of LJ-P and LJ-Q with 3' enhancer deletions(M-MuLV nt 7938 to 8114), which self-inactivate 5' and 3' LTR promoters on viral replication and proviral insertion (8) so that neotranscription is predominantly or completely from the internal SV40 promoters. LJ-PAdMLPEnh- and LJ-QAdMLPEnh- are identical toLJ-PEnh- ad LJ-QEnh- except that they use the AdMLP (nt -248 to +34) (11111) as their internal promoter.

responsible for EC cell repression, we constructed variousPBSQ vectors which contain either 47-bp (M-MuLV nt 139to 185) or 28-bp (M-MuLV nt 147 to 174) inserts, derivedfrom either the wt or B2 mutant RBS regions, at the uniqueBamHI site. neo titers, expressed as G418-resistant CFU permilliliter of viral supernatant, indicate the level of expressionof these retroviral constructs in EC and NIH 3T3 cells (Table1). The EC cell restriction (NIH 3T3 to EC titer) for each ofthe constructs shows that the wt 47- and 28-bp sequencesrestrict expression in EC cells when inserted in eitherorientation into the BamHI site of PBSQ, as indicated byratios of greater than 3,500. However, the ratio of NIH 3T3to EC titers observed for the PBSQ constructs containing themutant 47- and 28-bp inserts were 50- to 100-fold lower.Thus, the difference in EC cell restriction between PBSQ wtand mutant insert constructs is similar to the differencebetween MP10 and PBSQ itself.These results allow us to define a stem cell cis-acting RBS

element which maps to M-MuLV nt 147 to 174. EC cellrepression occurs when the RBS is inserted in either orien-tation and in an intron downstream of the M-MuLV LTRpromoter, suggesting that the repression occurs at the DNA,rather than the RNA, level. The position of the RBS down-stream of the viral LTR promoter in our constructs indicatesthat this activity can work at a distance. That this repressor

activity is independent of viral proteins has been shown byrepression in direct transfections of EC cells with MP10versus the B2 mutant (5) and with the constructs (Fig. 1)PBSQ-wt28- versus PBSQ-m28- (data not shown).RBS repression of heterologous promoters. To further

characterize the EC cell restriction properties of theM-MuLV RBS, we studied the effects of wt and PBSQ RBSsequences on the simian virus 40 (SV40) early promoter andadenovirus major late promoter (AdMLP). In these experi-ments, viral vectors expressing the neo gene from theinternal SV40 early promoter or AdMLP, with either wt orPBSQ RBS sequences in an upstream position, were com-pared (Fig. 1; Table 1). The LJ constructs, LJ-P and LJ-Q,are derived from the DOL vector described by Korman et al.(18). The neo gene is driven by the SV40 early promoter inthese constructs, and they are identical except that LJ-P hasa wt RBS region whereas LJ-Q has a PBSQ-like RBS region.The high NIH 3T3/EC ratio observed for LJ-P (Table 1; NIH3T3/EC ratio, 6,651) indicates that the wt RBS sequencerepressed neo expression from the SV40 promoter in ECcells relative to NIH 3T3 cells, when in an upstream posi-tion. The PBSQ variant, LJ-Q, showed a relatively low NIH3T3/EC ratio (Table 1; NIH 3T3/EC ratio, 80), indicating thatits RBS region did not mediate repression of the SV40 earlypromoter.

MOL. CELL. BIOL.

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/m

cb o

n 24

Jan

uary

202

2 by

119

.236

.85.

87.

STEM CELL-SPECIFIC SILENCER 1217

TABLE 1. Viral titers

Neomycin titers' in: Ratio of NIHConstruct' T/CttrNIH 3T3 cells EC cells 3T3/EC titers

MP1O 22,000 10 2,200PBSQ (expt 1) 12,000 660 18PBSQ (expt 2) 550,000 15,000 37

PBSQ-wt47+ 625,000 80 7,812PBSQ-wt47- 375,000 100 3,750PBSQ-m47+ 175,000 2,000 88PBSQ-m47- 250,000 20,000 13PBSQ-wt28+ 68,000 3 22,667PBSQ-wt28- 52,000 4 13,000PBSQ-m28+ 80,000 321 249PBSQ-m28- 82,000 675 121

LJ-P 286,000 43 6,651LJ-Q 272,000 3,400 80LJ-PEnh- 62,500 220 284LJ-QEnh- 50,000 2,650 19LJ-PAdMLPEnh- 7,800 20 390LJ-QAdMLPEnh- 18,500 2,050 9

a Constructs are as described in the legend to Fig. 1.b Titers in NIH 3T3 and EC cells for a given viral supernatant were

determined at the same time. neo titers are expressed as G418-resistant CFUper milliliter of viral supernatant. For this table, the EC cells used were F9cells, although similar results have been obtained with PCC4 cells. The ratioof NIH 3T3 to EC titers is given as an indicator of the EC cell restriction ofa particular viral construct: higher values are indicative of greater restrictionin EC cells. Note that results with two different PBSQ viral supernatants areshown to indicate that absolute titers (as defined by NIH 3T3 titers) may varyas a result of factors such as time of supematant collection or density ofvirus-producing cells, but the EC restriction value (NIH 3T3/EC) is relativelyinvariant (18 versus 37).

It could be argued that LJ-P and LJ-Q expression levelsare influenced by their wt M-MuLV LTR promoters. Toaddress this concern, we performed additional experimentswith the M-MuLV enhancer-deleted constructs LJ-PEnh-and LJ-QEnh- (Fig. 1). These enhancer deletions signifi-cantly reduced the level ofM-MuLV LTR-driven transcriptsin NIH 3T3 cells (Fig. 2, lanes C and D versus A and B) andeliminated the already low levels of LTR-driven transcriptsin EC cells (lane F versus lane E). When LJ-PEnh- andLJ-QEnh- were assayed and the NIH 3T3/EC ratios wereanalyzed, we still observed a 15-fold difference in EC cellrestriction between the two. The AdMLP constructs, U-PAdMLPEnh- (Table 1; NIH 3T3/EC ratio, 390) and U-QAdMLPEnh- (Table 1; NIH 3T3/EC ratio, 9), showed asimilar RBS effect. The EC cell repression observed forthese constructs was more than 40-fold greater in the wtconstruct than in the PBSQ construct. These findings sug-gest that the inhibitory effect on the M-MuLV LTR promoterattributed to the RBS, when in a downstream position, canbe expanded to include the heterologous SV40 early pro-moter and AdMLP when in an upstream position.

Specific binding of a cellular factor to the M-MuLV RBS.The results described above suggest that RBS-mediatedrepression occurs at the DNA level and that undifferentiatedEC cells possess a trans-acting DNA-binding factor whichbinds to the M-MuLV RBS to repress gene expression. Toidentify such a factor, we have used the gel electrophoresisDNA-binding assay (band shift or gel retardation assay [33,37]) with our wt and single-base-pair (B2) mutant 28-bp RBSsequences as probes.Nuclear extracts prepared from undifferentiated PCC4

A B C D E F

MuLV -*

SV40 -bs: WI -

FIG. 2. neo transcripts from M-MuLV and SV40 promoters ininfected cells. Total cellular RNA from infected G418-resistant NIH3T3 (lanes A to D) or EC (F9) (lanes E and F) cells was fractionatedby denaturing agarose gel electrophoresis and blotted onto nitrocel-lulose as described in Materials and Methods. Proviral neomycin(neo) transcripts were detected by autoradiography after hybridiza-tion with a radiolabeled neo probe. Lanes: A, 7 ,ug of RNA fromNIH 3T3 cells infected with LJ-P; B, 8.5 jig of RNA from NIH 3T3cells infected with LJ-Q; C, 10.5 ,ug of RNA from NIH 3T3 cellsinfected with LJ-PEnh-; D, 14.0 ,ug of RNA from NIH 3T3 cellsinfected with LJ-QEnh-; E, 14.0 ,ug of RNA from EC cells infectedwith LJ-Q; F, 26 jig ofRNA from EC cells infected with LJ-QEnh-.Constructs are as described in the legend to Fig. 1. Transcriptsinitiated by M-MuLV LTR and SV40 early promoters are as indi-cated. Note that no neo signal is detected with RNA from uninfectedNIH 3T3 or EC cells (data not shown). All lanes were fractionated onthe same gel and hybridized on the same filter, but lanes E and Frepresent threefold-longer exposures than lanes A to D.

cells possess at least five factors which bind to the wt 28-bpRBS probe (Fig. 3, lane 2, bands A, B, C, D, and E versusthe free probe, band F). Of these bands, only band A appearsto be specific for the wt RBS probe and is absent when themutant probe is used. The cellular factor responsible forband shift A was also present in cytoplasmic fractions ofPCC4 cells (lane 6 [wt] versus lane 15 [mutant]), as well asnuclear and cytoplasmic fractions of extracts from undiffer-entiated F9 EC cells (data not shown).To examine the specificity of binding for factor A, we

performed competition experiments by using unlabeled com-petitor fragments in our DNA-binding reactions (Fig. 3 and4). The addition of 4 ng or more of unlabeled double-stranded wt fragment effectively competed for the binding offactor A (Fig. 3, lanes 5 and 9; Fig. 4, lanes 2 and 8), whereasthe addition of unlabeled double-stranded mutant fragmentdid not compete for this binding (Fig. 4, lanes 3 and 9). Theaddition of several different unlabeled single-stranded com-petitor DNAs (ssDNA) to the binding reactions did notinterfere with factor A binding (Fig. 3, lanes 3, 4, 7, and 8;Fig. 4, lanes 4 to 6 and 10 to 12). As a control, it is clear thatssDNA did compete for binding of other factors in Fig. 3 and4 (see bands B and C).

Methylation and hemimethylation studies. Factor A bindingis very sensitive to changes in a variety of binding condi-tions. Because of this, we have not found it possible toisolate enough factor A complex to perform convincingDNase I protection or standard methylation interferenceassays, even when the reaction volumes were increased 5- to20-fold. To circumvent this problem, we are attempting thepartial purification of factor A for eventual use in suchstudies. However, we also have identified specific bindingcontacts by use of bacterial methylases. In these experi-ments we methylated our wt 28-bp RBS probe in vitro withdam, HpaII, and MspI methylases. Factor A binding wasnot affected by dam methylation of the A residues in GATCsequences on both strands of the wt 28-bp probe (data notshown). In contrast, factor A binding was abolished whenthe wt probe was methylated at the internal or externalcytosines of the sequence CCGG by HpaII and MspI meth-

VOL . 1 l, 1991

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/m

cb o

n 24

Jan

uary

202

2 by

119

.236

.85.

87.

1218 PETERSEN ET AL.

2_ 3 4 5 61 9 11½' 41t5

.4$vkwtw"t~~~~~~~~~~~~~~~~~~~~~~~~~i

c~~~i W~~&A.WA8-Be+

-0 .

CM+

Dm4 *"* .0

En+ s

F-IFIG. 3. M-MuLV RBS binding factors. The double-stranded (ds)

wt or mutant (m) 28-bp probes below were labeled and isolated as

described in Materials and Methods.

wt5'GGGGG CTCGT CCGGG ATCGG GAGCA CCC3'3' CCCCC GAGCA GGCCC TAGCC CTCGT GGG 5'

m 5' GGGGG CTCGT CCGaG ATCGG GAGCA CCC 3'3' CCCCC GAGCA GGCtC TAGCC CTCGT GGG5'

We used 20,000 cpm of wt (lanes 1 to 9) or mutant (lanes 10 to 15)probe in each binding assay. The probes were incubated in 20 ,ulreactions with no extract (lanes 1 and 10), with 50 ,ug of PCC4 ECcell nuclear extract per ml (lanes 2 to 5 and 11 to 14) or with 250 ,ugof PCC4EC cell cytoplasmic extract per ml (lanes 6 to 9 and 15). Inthe following lanes, 4 ng of unlabeled competitor DNA was added:M13 ssDNA, lanes 3, 7, and 12; ssDNA wt probe, lanes 4, 8, and 13;dsDNA wt probe, lanes 5, 9, and 14. Incubations of 15 min at 25°Cwere performed by using standard binding conditions (37) plus 100ng of dI-dC, 5 mM KCI, and 5 mM NaCl. Reactions were terminatedby addition of loading dye, and free and complexed probes wereseparated by electrophoresis on a 6% acrylamide-Tris-glycine gel.Bands visualized after autoradiography with intensifying screes arecomplexes A, B, C, D, and E, as well as free probe, F. Band A isspecifically detected with the wt but not the mutant probe. Similarresults are obtained when probes are labeled on the opposite strand.Note that band B is variable in our hands and is not detected in allgels.

ylases, respectively (Fig. 5, lanes 4 and 7 [HpaII] and 5 and8 [MspI] versus lanes 2 and 3 [unmethylated] and 6 and 9[mock methylated]). Interestingly, although methylation in-terfered with factor A binding, we also observed a new

DNA-binding factor (designated Hp) that is detected withthe HpaII methylated probe (Fig. 5, lanes 4 and 7).

FIG. 4. Binding-factor competition studies. Wt probe, preparedas described in the legend to Fig. 3, was incubated under standardconditions with 50 ,ug of PCC4EC cell extract per ml. In lanes 2 to6, 4 ng of unlabeled competitor DNA was added; in lanes 8 to 12, 10ng of competitor DNA was added. Competitor DNAs were asfollows: wt dsDNA probe, lanes 2 and 8; mutant dsDNA, lanes 3and 9; wt sense strand ssDNA, lanes 4 and 10; wt antisense strandssDNA, lanes 5 and 11; M13 ssDNA, lanes 6 and 12. Bands A andC correspond to bands A and C in Fig. 3. Free probe is indicated (F)at the bottom of the gel.

To identify specific nucleotides involved in binding, wealso performed band shift assays with hemimethylated formsof the wt probe. For these experiments, labeled wt probeswere methylated; denatured; renatured with an excess ofunmethylated, unlabeled complementary strand; and reiso-lated. Factor A binding was unaffected when the sensestrand was methylated with HpaII (Fig. 5, lane 10) or MspI(lane 11) methylases, but was drastically diminished whenthe antisense strand was methylated with HpaII (lane 12) orMspI (lane 13) methylases. (Note that the reduction of factorA complex with hemimethylated probes in lanes 12 and 13was comparable to the effect of total methylation in lanes 4,5, 7, and 8.) Surprisingly, formation of the factor Hpcomplex, which requires CpG methylation (lanes 4 and 7),was dependent on methylation on the wt probe sense strand(lane 10) and did not occur when only the antisense CpG was

cytosine methylated (lane 12). Thus, both Hp and A factorsbind to the sense strand hemimethylated probes, but neitherbind probe which is antisense hemimethylated. To verify

2 3 4 '"b .8 9 1011 'd

AV

.? rY,

.i'

.,:,. I

MOL. CELL. BIOL.

k 0 'm '. .,O

Z.:i. i,

.lwimwip 4 %

im-AliNk..,:-w-Aw.11%,

.:, W*-

v4...

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/m

cb o

n 24

Jan

uary

202

2 by

119

.236

.85.

87.

STEM CELL-SPECIFIC SILENCER 1219

Hp

M WT Hp 'Ms 'VWT HP' Ms's WTfHp-' Ms' Hp'- Ms'-

FIG. 5. Factor binding to methylated and hemimethylated

probes. Probes were labeled as described in Materials and Methods

and incubated, as described in the legend to Fig. 3, with 175 p.g of

PCC4 extract per ml prior to electrophoretic separation. Probes

were as follows: lane 1, mutant probe; lane 2, wt sense strand

labeled; lane 3, wt antisense strand labeled; lane 4, wt fully HpaII

methylated, sense strand labeled; lane 5, wt fully Mspl methylated,

sense strand labeled; lane 6, wt mock methylated, sense strand

labeled; lane 7, wt fully Hpall methylated, antisense strand labeled;

lane 8, wt fully MspI methylated, antisense strand labeled; lane 9, wt

mock methylated, antisense strand labeled; lanes 10 to 13, wt probes

labeled on the sense (lanes 10 and 11) or antisense (lanes 12 and 13)

strands and hemimethylated on the lowercase nucleotide as shown

below:

lane 10, sense strand Hpall hemimethylated:

5' GGGGG CTCGT CcGGG ATCGG GAGAC CCC 3'

3' CCCCC GAGCA GGCCC TAGCC CTCTG GGG 5'

lanell, sense strand Mspl hemimethylated:

5' GGGGG CTCGT cCGGG ATCGG GAGAC CCC 3'

3' CCCCC GAGCA GGCCC TAGCCO CTCTG GGG 5'

lane 12, antisense strand Hpall hemimethylated:

5' GGGGG CTCGT CCGGG ATCGG GAGAC CCCY33' CCCCC GACCA GGcCC TAGCC CTCTG GGG 5'

lane 13, antisense strand MspI hemimethylated:

5' GGGGG CTCGT CCGGG ATCGG GAGAC CCCY33'COCCOC GAGCA GGCcC TAGCCO CTCTG GGG 5'

Methylation was checked as described in Materials and Methods.

Band A indicates binding factor A, and Hp designates the factor

which binds to the HpaII methylated wt probe. Only a portion of the

gel is shown.

this, we have excluded the trivial possibility that our an-

tisense HpalI probe contained nonspecific inhibitors of Hp

and A binding by performing a gel shift assay with combi-

nations of mixed probes (data not shown).

DISCUSSION

M-MuLV expression is restricted in EC cells. Although

virus integration occurs at normal levels (12, 35) steady-state

levels of RNA are decreased up to 100-fold in EC cells

relative to differentiated cell types. The retrovirus LTR

promoter functions poorly in EC cells, at least partly be-

cause EC cells appear to lack a positive trans-acting factor(s)

(20). Also, at least two cis-acting negative regulatory ele-ments participate in the EC repression phenomena, one ofwhich is in the vicinity of the primer-binding site (5, 11, 13,21, 22, 38).Some retroviral integrants are able to circumvent the EC

cell restriction by inserting downstream of strong cellularpromoters (5, 30), whereas others do so by mutation of oneof the cis-acting negative regulatory domains (5, 38). Ourevidence, and the evidence of others (21, 22), suggests thatthe M-MuLV RBS element mediates stem cell viral restric-tion by binding to an EC cell-specific trans-acting factorwhich interferes with expression. PBSQ, a recombinantretroviral construct containing mutations in the RBS region,is expressed at 100-fold-higher levels in undifferentiated ECcells than in wt M-MuLV constructs such as MP10, indicat-ing that repression is not simply a consequence of perfectPBS matching to a cellular tRNA. We have inserted wt andB2 host range mutant RBS sequences downstream of theviral LTR promoter in recombinant PBSQ constructs. Ourresults indicate that this 28-bp sequence can mediate repres-sion from the LTR promoter when placed in an intron ineither orientation (Fig. 1; Table 1).To analyze the effect of the RBS on heterologous promot-

ers, we placed our wt and mutant RBS elements upstream ofthe SV40 early promoter or the AdMLP driving neo expres-sion. In the constructs LJ-P, LJ-PEnh-, and LJ-PAdMLPEnh- (Fig. 1), the wt M-MuLV RBS sequence repressedEC cell expression from the SV40 early promoter andAdMLP (Table 1). In contrast, the variant RBS in LJ-Q,LJ-QEnh-, and LJ-QAdMLPEnh- did not repress expres-sion in EC cells. The enhancer-deleted constructs showedlittle or no neo expression from the viral LTR promoter as

assayed by Northern (RNA) blot analysis (Fig. 2). There-fore, we can conclude that the RBS effect observed for theseconstructs is on the SV40 early promoter and the AdMLP.Because we (5) and others (21) have observed RBS-

mediated EC cell-specific restriction of gene expression intransfections and have shown that such repression involvesthe regulation of RNA levels (22, 38), our evidence suggeststhat repression is mediated by an EC cell-specific regulatoryfactor which binds to the RBS at the DNA level. Our bandshift studies have identified a DNA-binding factor, factor A,which specifically binds to the wt RBS sequence and not tothe single-base-pair mutant sequence (Fig. 3 to 5). Loh et al.(23) also have reported the identification, by exonuclease IIIprotection analysis, of a factor which specifically binds tothe wt RBS sequence. The extreme sensitivity of factor Ahas precluded ready complex isolation for standard DNase Iprotection and methylation interference assays: it is hopedthat partial purification of the factor will facilitate suchstudies. Nevertheless, our analyses with methylated probes(Fig. 5) indicate several contact sites of factor A with the wtRBS. dam methylation at adenine nucleotides (M-MuLVsense nt 162 and antisense nt 163) did not impair factorbinding (data not shown), nor did cytosine methylation on

sense strand nt 157 and 158 (Fig. 5). However, cytosine C-5methylation on the M-MuLV antisense strand at nt 159 and160 drastically reduced complex A formation (Fig. 5). Thisresult implies that factor A binding involves major groovecontacts at these nucleotides. That factor A might mediateRBS repression is supported by the fact that MspI interfer-ence and RBS mutant studies indicate the importance ofM-MuLV nt 160 in their effects.Although the binding characteristics of factor A suggest

that it may be the RBS repressor, several observations are

problematic. We observe factor A binding in both nuclear

VOL . 1 l, 1991

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/m

cb o

n 24

Jan

uary

202

2 by

119

.236

.85.

87.

1220 PETERSEN ET AL.

and cytoplasmic extracts from undifferentiated EC cells,which may be the result of nuclear contamination of ourcytoplasmic extracts or may reflect the natural distributionof this factor. Although there is precedent for localization oftranscriptional factors in the cytoplasm (2, 3, 19), ourobservations are complicated by the fact that the half-life ofbinding factor A is 15 min in nuclear extracts but more than60 min in cytoplasmic extracts. Another observation whichrequires explanation is the identification of binding factor Aactivity in differentiated NIH 3T3 cells (data not shown).Because the RBS effect is not observed in NIH 3T3 cells, ourresults with factor A are not compatible with a simpleexplanatory model. One possibility is that RBS repressiondoes not involve factor A at all. However, it should be notedthat factor binding may be the first in a series of steps leadingto regulation.One last complication concerning factor A binding to the

RBS is the identification of factor Hp binding to the HpaIImethylated wt probe (Fig. 5). DNA methylation has beenshown to play a significant role in controlling gene expres-sion (35), yet the control of DNA methylation is very poorlyunderstood. Because EC cells demonstrate high levels ofCpG methylation (28, 35), it is possible that the Hp factor isthe functional RBS repressor and acts on methylated DNA,rather than repression mediated by factor A at unmethylatedsequences. In support of this scenario is the fact that Hp isunlike nonspecific CpG-binding proteins (1, 26), as demon-strated by its requirement for RBS sense strand methylation(Fig. 5). However, global methylation of M-MuLV in ECcells appears to take more than 7 days (28), and we haveobserved RBS repression at 2 days postinfection by usingluciferase reporter vectors (data not shown). Thus, unlessthe RBS CpG is methylated much more rapidly than the bulkof the M-MuLV DNA (a possibility that we are testing), itseems unlikely that initial M-MuLV repression in EC cellsinvolves the Hp factor. Nevertheless, we consider it possiblethat factors A and Hp participate together in RBS regulation,with A and Hp repressing expression from unmethylated andmethylated sequences, respectively. With this in mind, it isof interest that although neither Hp nor factor A binds toantisense HpaII hemimethylated probes, both factors bindto the wt sense hemimethylated probe. Therefore, immedi-ately after replication of fully methylated DNA, daughterstrands may differentially bind Hp and/or A factors. Thisfeature could be used to give daughter cells different devel-opmental potentials.We do not know whether RBS homologs are used in the

normal regulation of cellular genes in EC cells. TheM-MuLV RBS sequence contains no obvious identities withother nuclear-binding factor sites (16). However, in additionto the expected homologies with a variety of retrovirusPBSs, the M-MuLV RBS contains an unexpected homologyto the 3'-noncoding region of the alpha chain of the T-cellreceptor (TCR) (31):

M-MuLV repressor site:5'GGGGG CTCGT CCGGG ATCGG GAGAC CCC 3'

TCR 3' non-coding region:5'cGGGG CTtcT CCtGG ATCtG aAGAC CCC 3'

This homology is interesting because it maps to the TCR-alpha region that has been implicated in silencing TCR-alphagene expression in immature T cells (39). We do not knowthe significance, if any, of this similarity, because we havenot examined whether the above sequence differences affectRBS function. Nevertheless, since M-MuLV is a T-cell

leukemia virus, it is possible that the M-MuLV RBS plays arole in the regulation of the virus gene expression duringT-cell differentiation. We are studying the expression of thewt and m viral constructs in immature and mature T cells toascertain the functional significance of this homology.

ACKNOWLEDGMENTS

We are indebted to Nancy Speck and Nancy Manley for adviceand assistance concerning protocols and procedures. We thankRichard Scott for the communication of unpublished results. Thanksalso to Scott Landfear, Jorge Crosa, and David Kabat for helpfuland informative discussions.

This work was supported by grant MV416A from the AmericanCancer Society (National Chapter).

REFERENCES1. Antequara, F., D. Macleod, and A. P. Bird. 1989. Specific

protection of methylated CpGs in mammalian nuclei. Cell58:509-517.

2. Baeuerle, P. A., and D. Baltimnore. 1988. Activation of DNA-binding activity in an apparently cytoplasmic precursor of theNF-K, transcription factor. Cell 53:211-217.

3. Baeuerle, P. A., and D. Baltimore. 1988. IK: A specific inhibitorof the NF-K13 transcription factor. Science 242:540-546.

4. Barklis, E., and H. Lodish. 1983. Regulation of dictyosteliumdiscoideum mRNAs specific for prespore or prestalk cells. Cell32:1139-1148.

5. Barklis, E., R. C. Mulligan, and R. Jaenisch. 1986. Chromo-somal position or virus mutation permits retrovirus expressionin embryonal carcinoma cells. Cell 47:391-399.

6. Bernstine, E., M. Hooper, S. Grandchamp, and B. Ephrussi.1973. Alkaline phosphatase activity in mouse teratoma. Proc.Natl. Acad. Sci. USA 70:3899-3903.

7. Colicelli, J., and S. Goff. 1987. Isolation of a recombinantmurine leukemia virus utilizing a new primer tRNA. J. Virol.57:37-45.

8. Cone, R., A. Weber-Benarous, D. Baorto, and R. Mulligan. 1987.Regulated expression of a complete human 3-globin gene en-coded by a transmissible retrovirus vector. Mol. Cell. Biol.7:887-897.

9. Dignam, J. D., R. M. Lebovitz, and R. G. Roeder. 1983.Accurate transcription initiation by RNA polymerase II in asoluble extract from isolated mammalian nuclei. Nucleic AcidsRes. 11:1475-1489.

10. Evans, M. J., and M. H. Kaufman. 1981. Establishment inculture of pluripotent cells from mouse embryos. Nature (Lon-don) 292:154-156.

11. Feuer, F., M. Taketo, R. C. Hanecak, and H. Fan. 1989. Twoblocks in Moloney murine leukemia virus expression in undif-ferentiated F9 embryonal carcinoma cells as determined bytransient expression assays. J. Virol. 63:2317-2324.

12. Gautsch, J., and M. Wilson. 1983. Restriction of moloneymurine leukemia virus growth in teratocarcinoma: involvementof factors other than DNA methylation. Cold Spring HarborConf. Cell Proliferation 10:363-378.

13. Gorman, C., P. Rigby, and D. Lane. 1985. Negative regulationof viral enhances in undifferentiated embryonic stem cells. Cell42:519-526.

14. Graham, F. L., and A. J. van der Eb. 1973. A new technique forthe assay of infectivity of human adenovirus 5 DNA. Virology52:456-467.

15. Jaenisch, R., and A. Berns. 1977. Tumor virus expression duringmammalian embryogenesis, p. 267-314. In M. Sherman (ed.),Concepts in mammalian embryogenesis. Massachusetts Insti-tute of Technology Press, Cambridge, Mass.

16. Jones, N. C., P. W. J. Rigby, and E. B. Ziff. 1988. Trans-actingprotein factors and the regulation of eukaryotic transcriptionlessons from studies on DNA tumor viruses. Genes Dev.2:267-281.

17. Jones, T. A., G. Blaug, M. Hansen, and E. Barklis. 1990.Assembly of gag-p-galactosidase proteins into retrovirus parti-cles. J. Virol. 64:2265-2279.

MOL. CELL. BIOL.

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/m

cb o

n 24

Jan

uary

202

2 by

119

.236

.85.

87.

STEM CELL-SPECIFIC SILENCER 1221

18. Korman, A. J., J. D. Frantz, J. L. Strominger, and R. C.Mulligan. 1987. Expression of human class II major histocom-patibility complex antigens using retrovirus vectors. Proc. Natl.Acad. Sci. USA 84:2150-2154.

19. Levy, D. E., D. S. Kessler, R. Pine, and J. E. Darnell, Jr. 1989.Cytoplasmic activation of ISGF3, the positive regulator ofinterferon-a-stimulated transcription, reconstituted in vitro.Genes Dev. 3:1362-1371.

20. Linney, E., B. Davis, J. Overhauser, E. Chao, and H. Fan. 1984.Nonfunction of a moloney murine leukemia virus regulatorysequence in F9 embryonal carcinoma cells. Nature (London)308:470-472.

21. Loh, T. P., L. L. Sievert, and R. W. Scott. 1987. Proviralsequences that restrict retroviral expression in mouse embryo-nal carcinoma cells. Mol. Cell. Biol. 7:3775-3784.

22. Loh, T. P., L. L. Sievert, and R. W. Scott. 1988. Negativeregulation of retrovirus expression in embryonal carcinomacells mediated by an intragenic domain. J. Virol. 62:4086-4095.

23. Loh, T. P., L. L. Sievert, and R. W. Scott. 1990. Evidence for astem cell-specific repressor of Moloney murine leukemia virusexpression in embryonal carcinoma cells. Mol. Cell. Biol.10:4045-4057.

24. Maniatis, T., E. F. Fritsch, and J. Sambrook. 1982. Molecularcloning: a laboratory manual. Cold Spring Harbor Laboratory,Cold Spring Harbor, N.Y.

25. Mann, R., R. Mulligan, and D. Baltimore. 1983. Construction ofa retrovirus packaging mutant and its use to produce helper-freedefective retrovirus. Cell 33:153-159.

26. Meehan, R., J. D. Lewis, S. McKay, E. L. Kleiner, and A. P.Bird. 1989. Identification of a mammalian protein that bindsspecifically to DNA containing methylated CpGs. Cell 58:499-507.

27. Miller, A., and C. Buttimore. 1986. Redesign of retroviruspackaging cell lines to avoid recombination leading to helpervirus production. Mol. Cell. Biol. 6:2895-2902.

28. Niwa, O., Y. Yokota, H. Ishida, and T. Sugahara. 1983. Inde-

pendent mechanisms involved in suppression of the moloneyleukemia virus genome during differentiation of murine terato-carcinoma cells. Cell 32:1105-1113.

29. Parker, B., and G. Stark. 1979. Regulation of simian virus 40transcription: sensitive analysis of the RNA species presentearly in infection. J. Virol. 31:360-369.

30. Peckham, I., S. Sobel, J. Comer, R. Jaenisch, and E. Barklis.1989. Retrovirus activation in embryonal carcinoma cells bycellular promoters. Genes Dev. 3:2062-2071.

31. Saito, H., D. Kranz, Y. Takagaki, A. Hayday, H. Eisen, and S.Tonegawa. 1984. A third rearranged and expressed gene in aclone of cytotoxic T lymphocytes. Nature (London) 312:36-40.

32. Shinnick, T., R. Lerner, and J. Sutcliffe. 1981. Nucleotidesequence of Moloney murine leukemia virus. Nature (London)293:543-548.

33. Speck, N. A., and D. Baltimore. 1987. Six distinct nuclearfactors interact with the 75-base-pair repeat of moloney murineleukemia virus enhancer. Mol. Cell. Biol. 7:1101-1110.

34. Stevens, L. 1967. The biology of teratomas. Adv. Morphol. 6:1.35. Stewart, C., H. Stuhlman, D. Jahner, and R. Jaenisch. 1982. De

novo methylation, expression, and infectivity of retroviral ge-nomes introduced into embryonal carcinoma cells. Proc. Natl.Acad. Sci. USA 79:4088-4092.

36. Teich, N., R. Weiss, G. Martin, and D. Lowy. 1977. Virusinfection of murine teratocarcinoma stem cell lines. Cell 12:973-982.

37. Thornell, A., B. Hallberg, and T. Grundstrom. 1988. Differentialprotein binding in lymphocytes to a sequence in the enhancer ofthe mouse retrovirus SL3-3. Mol. Cell. Biol. 8:1625-1637.

38. Weiher, H., E. Barklis, W. Ostertag, and R. Jaenisch. 1987. Twodistinct sequence elements mediate retroviral gene expressionin embryonal carcinoma cells. J. Virol. 61:2742-2746.

39. Winoto, A., and D. Baltimore. 1989. au lineage-specific repres-sion of the a T cell receptor gene by nearby silencers. Cell59:649-655.

VOL . 1 l, 1991

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/m

cb o

n 24

Jan

uary

202

2 by

119

.236

.85.

87.